Information Technology Reference
In-Depth Information
Looking at the relationship between specific viewpoints, some follow a sin-
gle pattern, while others are composite in nature. The relationship between
the computational and engineering viewpoints is perhaps the simplest, be-
ing a straightforward refinement relation, in which the engineering viewpoint
provides a set of templates for refining the computational specification. This
results in an ephemeral specification that is not one of the ODP viewpoints,
but represents an abstract implementation. Since it will be regenerated afresh
whenever necessary, it will track changes in the computational and engineering
viewpoints as they occur.
The relationship between the information and computational viewpoints
involves a merge, in which information types are imported into the computa-
tional specification, where they will generally be used as a starting point for
further refinement as part of the computational design process.
The relationship between the enterprise specification and the others is
again primarily a refinement, but the rules to be applied depend on the orga-
nization and the methodology in use, so that many approaches to unification
are possible.
The involvement of the technology viewpoint is more complex because
it deals with several different kinds of information. The viewpoint covers
the available resources and their configuration, as well as catalogues of imple-
mentable standards and conformance requirements. One of the main activities
is the performance of an allocation step that associates tasks in the abstract
implementation to specific resources; the results of this allocating transforma-
tion is added to the technology viewpoint configuration information. There
is then an instance of relationship between technology viewpoint elements
produced and the engineering elements being supported.
These examples give an idea of how transformational tools can manipulate
viewpoint information and support the realization of the system that has been
designed. There are, of course, other information flows as part of the system
generation process, drawn from the patterns given previously, or otherwise,
but it should be clear that the tools will each need to act on and combine
information from different viewpoints to do their job.
Model transformations can also be very useful for keeping different models
synchronized. These models can be either two representations of the same
viewpoint specification, or two viewpoint specifications related by a set of
correspondences. The first case happens when the user applies two concrete
syntaxes to express one viewpoint language, for example a textual and a graph-
ical one. In the second case, changes in one of the viewpoint specifications
may need to be reflected in the rest of the related viewpoints. As indicated
in chapter 7, the correspondences can be used to propagate these changes,
and model transformations provide a natural mechanism for keeping these
specifications in step. For instance, think of two elements living in different
viewpoints, but related by a correspondence that establishes that their names
should coincide. Model transformations can be responsible for changing the
name of an element whenever the related one is modified.
